AI in Business Enterprises

Transcription

1 AI in Business Enterprises Are Humans Rational? Rini Palitmittam 10 th October 2017

2 Image Courtesy: Google Images Founders of Modern Artificial Intelligence

3 Image Courtesy: Google Images Founders of Modern Artificial Intelligence

4 Image Courtesy: Google Images Founders of Modern Artificial John McCarthy Intelligence

5 Image Courtesy: Google Images Founders of Modern Artificial John McCarthy Intelligence

6 Image Courtesy: Google Images Founders of Modern Artificial John McCarthy Intelligence Herb A Simon

7 Image Courtesy: Google Images Founders of Modern Artificial John McCarthy Intelligence Herb A Simon

8 Image Courtesy: Google Images Founders of Modern Artificial Intelligence John McCarthy Allen Newell Herb A Simon

9 Image Courtesy: Google Images Founders of Modern Artificial Intelligence John McCarthy Allen Newell Herb A Simon

10 Image Courtesy: Google Images Founders of Modern Artificial Intelligence John McCarthy Allen Newell Herb A Simon Marvin Minsky

11 Image Courtesy: Google Images Founders of Modern Artificial Intelligence John McCarthy Allen Newell - Mathematician/ Computer Scientist - Developed Lisp - Turing Award, National Medal Science and many more.. Herb A Simon Marvin Minsky

12 Founders of Modern Artificial Intelligence John McCarthy Allen Newell - Mathematician/ Computer Scientist - Developed Lisp - Turing Award, National Medal Science and many more.. Herb A Simon -Economist, Computer & Cognitive Scientist, Psychologist. Marvin Minsky -Work in Problem solving, decision-making, organization theory, simulations in scientific discovery -Nobel Prize in Economics & Turing Award Image Courtesy: Google Images

13 Founders of Modern Artificial Intelligence John McCarthy - Mathematician/ Computer Scientist - Developed Lisp - Turing Award, National Medal Science and many more.. Allen Newell - Computer Scientist, Cognitive Psychologist - Developed Information Processing Language. - Turing Award, National Medal Science and many more.. Herb A Simon -Economist, Computer & Cognitive Scientist, Psychologist. Marvin Minsky -Work in Problem solving, decision-making, organization theory, simulations in scientific discovery -Nobel Prize in Economics & Turing Award Image Courtesy: Google Images

14 Founders of Modern Artificial Intelligence John McCarthy - Mathematician/ Computer Scientist - Developed Lisp - Turing Award, National Medal Science and many more.. Allen Newell - Computer Scientist, Cognitive Psychologist - Developed Information Processing Language. - Turing Award, National Medal Science and many more.. Herb A Simon -Economist, Computer & Cognitive Scientist, Psychologist. -Work in Problem solving, decision-making, organization theory, simulations in scientific discovery Marvin Minsky - Cognitive Scientist -Neural Networks - Turing Award, Benjamin Franklin Medal and many more.. -Nobel Prize in Economics & Turing Award Image Courtesy: Google Images

15 Evolution of AI over the years!

16 Evolution of AI over the years! 1949

17 Evolution of AI over the years!

18 Evolution of AI over the years!

19 Evolution of AI over the years!

20 Evolution of AI over the years!

21 Evolution of AI over the years!

22 Evolution of AI over the years!

23 Evolution of AI over the years! 1949 Alan Turing comes up with a Test, today known as Turing Test

24 Evolution of AI over the years! 1949 Alan Turing comes up with a Test, today known as Turing Test 1952 A computer program developed by Arthur Samuel for Checkers Game

25 Evolution of AI over the years! 1949 Alan Turing comes up with a Test, today known as Turing Test A computer program developed by Arthur Samuel for Checkers Game The term Artificial Intelligence was coined by John McCarthy

26 Evolution of AI over the years! 1949 Alan Turing comes up with a Test, today known as Turing Test A computer program developed by Arthur Samuel for Checkers Game The term Artificial Intelligence was coined by John McCarthy The Logic Theorist developed by Allen Newell & Herb Simon

27 Evolution of AI over the years! 1949 Alan Turing comes up with a Test, today known as Turing Test A computer program developed by Arthur Samuel for Checkers Game The term Artificial Intelligence was coined by John McCarthy The Logic Theorist developed by Allen Newell & Herb Simon 1957 The General Problem Solver developed by Allen Newell & Herb Simon

28 Evolution of AI over the years! 1949 Alan Turing comes up with a Test, today known as Turing Test A computer program developed by Arthur Samuel for Checkers Game The term Artificial Intelligence was coined by John McCarthy The Logic Theorist developed by Allen Newell & Herb Simon The General Problem Solver developed by Allen Newell & Herb Simon Lisp developed by John McCarthy, Father of Modern AI. 1964

29 Evolution of AI over the years! 1949 Alan Turing comes up with a Test, today known as Turing Test A computer program developed by Arthur Samuel for Checkers Game The term Artificial Intelligence was coined by John McCarthy The Logic Theorist developed by Allen Newell & Herb Simon The General Problem Solver developed by Allen Newell & Herb Simon Lisp developed by John McCarthy, Father of Modern AI The year of Natural Language Processing. MIT s Daniel Bobrow s Dissertation Natural Language Input for a Computer Problem Solving System

30

31 Let us now get into the crux of the lecture today..

32

33 Artificial Intelligence in Business Enterprises

34

35 So, what happened after 1964? Any Guesses?

36 So, what happened after 1964? Any Guesses? Continuing on with the timeline.

37

38 The Emergence of Expert Systems

39 The Emergence of Expert Systems 1965

40 The Emergence of Expert Systems

41 The Emergence of Expert Systems

42 The Emergence of Expert Systems

43 The Emergence of Expert Systems

44 The Emergence of Expert Systems Early 1980 s 1987

45 The Emergence of Expert Systems Early 1980 s

46 The Emergence of Expert Systems 1965 A Stanford Project DENDRAL : the first expert system developed to the aid of Organic Chemists Early 1980 s

47 The Emergence of Expert Systems A Stanford Project DENDRAL : the first expert system developed to the aid of Organic Chemists. The ERA of EXPERT SYSTEMS Development of an early expert systems such as MYCIN at Stanford. Most Expert Systems were built in LISP Early 1980 s

48 The Emergence of Expert Systems A Stanford Project DENDRAL : the first expert system developed to the aid of Organic Chemists. The ERA of EXPERT SYSTEMS Development of an early expert systems such as MYCIN at Stanford. Most Expert Systems were built in LISP. DECLINE OF LISP, ENTER PROLOG! 1974 Early 1980 s

49 The Emergence of Expert Systems Early 1980 s A Stanford Project DENDRAL : the first expert system developed to the aid of Organic Chemists. The ERA of EXPERT SYSTEMS Development of an early expert systems such as MYCIN at Stanford. Most Expert Systems were built in LISP. DECLINE OF LISP, ENTER PROLOG! The Onset of AI WINTER Some of the expert systems were still in use during this time

50 The Emergence of Expert Systems Early 1980 s A Stanford Project DENDRAL : the first expert system developed to the aid of Organic Chemists. The ERA of EXPERT SYSTEMS Development of an early expert systems such as MYCIN at Stanford. Most Expert Systems were built in LISP. DECLINE OF LISP, ENTER PROLOG! The Onset of AI WINTER Some of the expert systems were still in use during this time The Onset of AI WINTER-2 Some of the expert systems were still in use during this time.

51

52 WORK OF THE AI PIONEERS

53 WORK OF THE AI PIONEERS! Logic Theorist (in collaboration with Allen Newell and Cliff Shaw):! Used Search Trees.(Finding the goal node)! Introduced the concept of Heuristics (Ad-hoc Rules)! Implementation in a Programming Language called IPL that was based on Symbolic List Processing and this was in Assemble Language style.

54 WORK OF THE AI PIONEERS! Logic Theorist (in collaboration with Allen Newell and Cliff Shaw):! Used Search Trees.(Finding the goal node)! Introduced the concept of Heuristics (Ad-hoc Rules)! Implementation in a Programming Language called IPL that was based on Symbolic List Processing and this was in Assemble Language style.! The General Problem Solver (in collaboration with Allen Newell and Cliff Shaw):! A Universal problem solver machine.! Anything that can expressed as well-formed formula or in a directed graph form can be solved in GPS! Predicate Logic proofs could be easily solved.! GPS, then evolved into SOAR architecture(a Cognitive Architecture for Intelligent Agents)

55 WORK OF THE AI PIONEERS! Logic Theorist (in collaboration with Allen Newell and Cliff Shaw):! Used Search Trees.(Finding the goal node)! Introduced the concept of Heuristics (Ad-hoc Rules)! Implementation in a Programming Language called IPL that was based on Symbolic List Processing and this was in Assemble Language style.! The General Problem Solver (in collaboration with Allen Newell and Cliff Shaw):! A Universal problem solver machine.! Anything that can expressed as well-formed formula or in a directed graph form can be solved in GPS! Predicate Logic proofs could be easily solved.! GPS, then evolved into SOAR architecture(a Cognitive Architecture for Intelligent Agents)! BUT MORE THE GENERALIZATION, MORE IS THE COMPUTATIONAL STRESS!

56

57 NOTICE SIMON S POSITIVITY TOWARD COMPUTATIONAL LOGIC?

58 NOTICE SIMON S POSITIVITY TOWARD COMPUTATIONAL LOGIC? NOTICE THE DRASTIC CHANGE IN ARTIFICIAL INTELLIGENCE APPROACHES?

59

60 What happened to the vision of our Pioneers? How are today s systems different from what the Pioneers of AI had in mind?

61 What happened to the vision of our Pioneers? How are today s systems different from what the Pioneers of AI had in mind?! Focus has shifted Statistical Machine Learning, Neural Networks, Deep Learning

62 What happened to the vision of our Pioneers? How are today s systems different from what the Pioneers of AI had in mind?! Focus has shifted Statistical Machine Learning, Neural Networks, Deep Learning! They might be popular today, however they have their disadvantages.

63 What happened to the vision of our Pioneers? How are today s systems different from what the Pioneers of AI had in mind?! Focus has shifted Statistical Machine Learning, Neural Networks, Deep Learning! They might be popular today, however they have their disadvantages.! CAN YOU THINK OF THEIR LIMITATIONS OR DISADVANTAGES?

64 Modern Day Enterprise Systems

65 Modern Day Enterprise Systems Transaction Processing Systems

66 Modern Day Enterprise Systems Management Information Systems Transaction Processing Systems

67 Modern Day Enterprise Systems Decision Support Systems Management Information Systems Transaction Processing Systems

68 Modern Day Enterprise Systems Executive Information Systems Decision Support Systems Management Information Systems Transaction Processing Systems

69 Modern Day Enterprise Systems Executive Information Systems Decision Support Systems Management Information Systems Captures daily transactional information like purchase orders, delivery information, etc. Transaction Processing Systems

70 Modern Day Enterprise Systems Executive Information Systems Decision Support Systems Collecting, cleaning, transforming and managing the information collected from TPS. Captures daily transactional information like purchase orders, delivery information, etc. Management Information Systems Transaction Processing Systems

71 Modern Day Enterprise Systems Executive Information Systems Analyzing the information from the MIS to gain insights into the data, predict outcomes, come up with decisions for future actions. Collecting, cleaning, transforming and managing the information collected from TPS. Captures daily transactional information like purchase orders, delivery information, etc. Decision Support Systems Management Information Systems Transaction Processing Systems

72 Modern Day Enterprise Systems Today known as Business Intelligence (dealing with Reporting, Dashboards, Analytics) Executive Information Systems Analyzing the information from the MIS to gain insights into the data, predict outcomes, come up with decisions for future actions. Collecting, cleaning, transforming and managing the information collected from TPS. Captures daily transactional information like purchase orders, delivery information, etc. Decision Support Systems Management Information Systems Transaction Processing Systems

73 Transactional Processes Captured Today s Enterprise systems are extremely focused on the transactions taking place between the Enterprise itself and external parties (Vendors, Suppliers, Partners etc.), transactions among the various departments of the company by capturing the information flow and finally the transactions between the Clients/ Customers and the Enterprise. Finance Transporation & Delivery Suppliers Manufacturing Marketing Vendors Customers SALES Information Tech. Partners

74

75 What about Internal Processes?

76 What about Internal Processes?! Very limited AI focus in this area.

77 What about Internal Processes?! Very limited AI focus in this area.! Extremely complex.

78 What about Internal Processes?! Very limited AI focus in this area.! Extremely complex.! Understanding of human behavior and psychology, organizational theory, decision-making, planning, goals.

79 What about Internal Processes?! Very limited AI focus in this area.! Extremely complex.! Understanding of human behavior and psychology, organizational theory, decision-making, planning, goals.! Degree of rationality involved in Choices and Decisions

80 What about Internal Processes?! Very limited AI focus in this area.! Extremely complex.! Understanding of human behavior and psychology, organizational theory, decision-making, planning, goals.! Degree of rationality involved in Choices and Decisions! Ethical Issues.

81 What about Internal Processes?! Very limited AI focus in this area.! Extremely complex.! Understanding of human behavior and psychology, organizational theory, decision-making, planning, goals.! Degree of rationality involved in Choices and Decisions! Ethical Issues.! Understanding company culture.

82

83 AND THIS IS WHERE HERB SIMON s WORK in ORGANZATIONAL THEORY COMES INTO PLAY

84 AND THIS IS WHERE HERB SIMON s WORK in ORGANZATIONAL THEORY COMES INTO PLAY! In this class, we will cover some areas that Herb Simon talks about in his work Administrative Behavior :! Means- End Schema! Knowledge and Behavior! A General View of Rationality! Touch upon Bounded Rationality

85 Means- Ends Analysis To understand what we are getting into next:

86

87 Means-End Approach

88 Means-End Approach! Means: Alternatives, possibilities

89 Means-End Approach! Means: Alternatives, possibilities! Ends: Goals/ Sub Goals

90 Means-End Approach! Means: Alternatives, possibilities! Ends: Goals/ Sub Goals! Sometimes, the difference between the means and the ends could be obscure.

91 Means-End Approach! Means: Alternatives, possibilities! Ends: Goals/ Sub Goals! Sometimes, the difference between the means and the ends could be obscure.! Policies play an an important role in this schema. Why?

92 Means-End Approach! Means: Alternatives, possibilities! Ends: Goals/ Sub Goals! Sometimes, the difference between the means and the ends could be obscure.! Policies play an an important role in this schema. Why?! For e.g.: For the fire department! End : Reducing fire losses! Means: Prevention of Fire & Extinguishing Fires

93

94 Limitations of Means End Schema:

95 Limitations of Means End Schema:! Compare all alternatives to achieve the goal.! Establishing sub-goals to achieve the long term, in the absence of knowledge of alternatives of how to achieve the goal.

96 Limitations of Means End Schema:! Compare all alternatives to achieve the goal.! Establishing sub-goals to achieve the long term, in the absence of knowledge of alternatives of how to achieve the goal.! Time Factor: Some decisions are irrevocable as new situations/ consequences arise once they are made.! Given a time constraint what alternative can be used?

97 Limitations of Means End Schema:! Compare all alternatives to achieve the goal.! Establishing sub-goals to achieve the long term, in the absence of knowledge of alternatives of how to achieve the goal.! Time Factor: Some decisions are irrevocable as new situations/ consequences arise once they are made.! Given a time constraint what alternative can be used?! A complete separation of means and ends is impossible. It may lead to having to deal with opposite consequences.! For e.g. Prohibition Amendment of 1919.

98 ! Decision is the process by which the most feasible and economic alternative is picked in order to achieve the goal.! Following are the steps involved in Decision Making:!List out all alternative Strategies.!Determination of all consequences that follow upon each of the strategies!comparative evaluation of these sets of consequences.

99 Decision-Making Theory! Decision is the process by which the most feasible and economic alternative is picked in order to achieve the goal.! Following are the steps involved in Decision Making:!List out all alternative Strategies.!Determination of all consequences that follow upon each of the strategies!comparative evaluation of these sets of consequences.

100

101 Role of Knowledge in Decision-Making

102 Role of Knowledge in Decision-Making! Determining what consequences follow after which consequences.

103 Role of Knowledge in Decision-Making! Determining what consequences follow after which consequences.! Selection from a whole class of consequences a smaller and limited subclass or in

104 Role of Knowledge in Decision-Making! Determining what consequences follow after which consequences.! Selection from a whole class of consequences a smaller and limited subclass or in! An ideal case a single set of consequences correlated with the strategy opted for.

105 Role of Knowledge in Decision-Making! Determining what consequences follow after which consequences.! Selection from a whole class of consequences a smaller and limited subclass or in! An ideal case a single set of consequences correlated with the strategy opted for.! For e.g. A Hiring Process in a Company

106

107 Concept of Bounded Rationality

108 Concept of Bounded Rationality! It is impossible for an individual to know all its alternatives or all their consequences because:! We have cognitive limitations of mind.! Impossible to have all knowledge about all strategies & alternatives.! Impossible to know of all the consequences.! Time constraint(finite time) to make the decision.

109

110 Concept of Bounded Rationality

111 Concept of Bounded Rationality! According to Herb Simon s Theory of Bounded Rationality,! Decision-makes tend to be satisficers! Satisficers are satisfied with a Satisfactory solution rather than an optimal.! HERB SIMON views decision-making as fully rational process of finding an optimal choice using the the information available to the individual.

112

113 Thank You!